1 core keeps running at 100% CPU after workflow completion

### Custom Node Testing

- [ ] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)

### Expected Behavior

Once the Wan 2.2 I2V workflow has finished, it should no longer cause a significant CPU load.

### Actual Behavior

1 core remains at 100% utilitzation.

### Steps to Reproduce

Run the flow in attachment, wait for it to finish, close the browser (but keep ComfyUI running), observe the CPU utilisation.

[Benchmark_I2V_wan22.json](https://github.com/user-attachments/files/30784737/Benchmark_I2V_wan22.json)

### Debug Logs

```powershell
## ComfyUI-Manager: installing dependencies done.
[2026-08-06 13:01:45.608] ** ComfyUI startup time: 2026-08-06 13:01:45.608
[2026-08-06 13:01:45.608] ** Platform: Linux
[2026-08-06 13:01:45.608] ** Python version: 3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]
[2026-08-06 13:01:45.608] ** Python executable: ~/comfyui-venv_7-14/bin/python3
[2026-08-06 13:01:45.608] ** ComfyUI Path: ~/ComfyUI
[2026-08-06 13:01:45.608] ** ComfyUI Base Folder Path: ~/ComfyUI
[2026-08-06 13:01:45.608] ** User directory: ~/ComfyUI/user
[2026-08-06 13:01:45.608] ** ComfyUI-Manager config path: ~/ComfyUI/user/__manager/config.ini
[2026-08-06 13:01:45.608] ** Log path: ~/ComfyUI/user/comfyui.log

Prestartup times for custom nodes:
[2026-08-06 13:01:45.668] 0.0 seconds: ~/ComfyUI/custom_nodes/rgthree-comfy
[2026-08-06 13:01:45.668] 0.0 seconds: ~/ComfyUI/custom_nodes/comfyui-easy-use
[2026-08-06 13:01:45.668] 0.3 seconds: ~/ComfyUI/custom_nodes/ComfyUI-Manager
[2026-08-06 13:01:45.668]
[2026-08-06 13:01:47.231] Found comfy_kitchen backend cuda: {'available': True, 'disabled': True, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope1_', 'apply_rope_', 'apply_rope_split_half', 'apply_rope_split_half1', 'apply_rope_split_half1_', 'apply_rope_split_half_', 'convrot_w4a4_linear', 'dequantize_convrot_w4a4_weight', 'dequantize_int8_convrot_weight', 'dequantize_int8_convrot_weight_dtype', 'dequantize_int8_simple', 'dequantize_int8_simple_dtype', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'gemv_awq_w4a16', 'prepare_int4_weight_for_int8_linear', 'quantize_and_rotate_rowwise', 'quantize_convrot_w4a4_weight', 'quantize_int8_convrot_weight', 'quantize_int8_rowwise', 'quantize_int8_tensorwise', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'quantize_svdquant_w4a4', 'rms_adaln', 'rms_rope', 'rms_rope1', 'rms_rope1_', 'rms_rope_', 'rms_rope_split_half', 'rms_rope_split_half1', 'rms_rope_split_half1_', 'rms_rope_split_half_', 'scaled_mm_svdquant_w4a4', 'stochastic_rounding_fp8']}
[2026-08-06 13:01:47.231] Found comfy_kitchen backend hip: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope1_', 'apply_rope_', 'apply_rope_split_half', 'apply_rope_split_half1', 'apply_rope_split_half1_', 'apply_rope_split_half_', 'convrot_w4a4_linear', 'dequantize_convrot_w4a4_weight', 'dequantize_int8_convrot_weight_dtype', 'dequantize_int8_simple_dtype', 'dequantize_per_tensor_fp8', 'gemv_awq_w4a16', 'int8_linear', 'quantize_and_rotate_rowwise', 'quantize_convrot_w4a4_weight', 'quantize_int8_convrot_weight', 'quantize_int8_rowwise', 'quantize_int8_tensorwise', 'quantize_per_tensor_fp8', 'quantize_svdquant_w4a4', 'rms_adaln', 'rms_rope', 'rms_rope1', 'rms_rope1_', 'rms_rope_', 'rms_rope_split_half', 'rms_rope_split_half1', 'rms_rope_split_half1_', 'rms_rope_split_half_', 'scaled_mm_svdquant_w4a4', 'stochastic_rounding_fp8']}
[2026-08-06 13:01:47.231] Found comfy_kitchen backend triton: {'available': True, 'disabled': True, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope1_', 'apply_rope_', 'apply_rope_split_half', 'apply_rope_split_half1', 'apply_rope_split_half1_', 'apply_rope_split_half_', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'int8_linear', 'quantize_and_rotate_rowwise', 'quantize_int8_rowwise', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'rms_adaln', 'rms_rope', 'rms_rope1', 'rms_rope1_', 'rms_rope_', 'rms_rope_split_half', 'rms_rope_split_half1', 'rms_rope_split_half1_', 'rms_rope_split_half_']}
[2026-08-06 13:01:47.231] Found comfy_kitchen backend eager: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope1_', 'apply_rope_', 'apply_rope_split_half', 'apply_rope_split_half1', 'apply_rope_split_half1_', 'apply_rope_split_half_', 'convrot_w4a4_linear', 'dequantize_convrot_w4a4_weight', 'dequantize_int8_convrot_weight', 'dequantize_int8_convrot_weight_dtype', 'dequantize_int8_embedding', 'dequantize_int8_simple', 'dequantize_int8_simple_dtype', 'dequantize_mxfp8', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'gemv_awq_w4a16', 'int8_linear', 'prepare_int4_weight_for_int8_linear', 'quantize_and_rotate_rowwise', 'quantize_convrot_w4a4_weight', 'quantize_int8_convrot_weight', 'quantize_int8_rowwise', 'quantize_int8_tensorwise', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'quantize_svdquant_w4a4', 'rms_adaln', 'rms_rope', 'rms_rope1', 'rms_rope1_', 'rms_rope_', 'rms_rope_split_half', 'rms_rope_split_half1', 'rms_rope_split_half1_', 'rms_rope_split_half_', 'scaled_mm_mxfp8', 'scaled_mm_nvfp4', 'scaled_mm_svdquant_w4a4', 'stochastic_rounding_fp8']}
[2026-08-06 13:01:47.233] Checkpoint files will always be loaded safely.
[2026-08-06 13:01:47.238] Total VRAM 24560 MB, total RAM 63934 MB
[2026-08-06 13:01:47.238] pytorch version: 2.12.0+rocm7.14.0
[2026-08-06 13:01:47.238] Set: torch.backends.cudnn.enabled = False for better AMD performance.
[2026-08-06 13:01:47.238] AMD arch: gfx1100
[2026-08-06 13:01:47.238] ROCm version: (7, 14)
[2026-08-06 13:01:47.239] Set vram state to: NORMAL_VRAM
[2026-08-06 13:01:47.239] Device: cuda:0 AMD Radeon RX 7900 XTX : native
[2026-08-06 13:01:47.239] Using async weight offloading with 2 streams
[2026-08-06 13:01:47.239] Enabled pinned memory 49597
[2026-08-06 13:01:47.330] Using Flash Attention
[2026-08-06 13:01:48.511] Python version: 3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]
[2026-08-06 13:01:48.511] ComfyUI version: 0.30.0
[2026-08-06 13:01:48.512] comfy-aimdo version: 0.4.13
[2026-08-06 13:01:48.512] comfy-kitchen version: 0.2.26
[2026-08-06 13:01:48.513] comfyui-frontend-package version: 1.48.6
[2026-08-06 13:01:48.513] comfyui-workflow-templates version: 0.11.31
[2026-08-06 13:01:48.513] comfyui-embedded-docs version: 0.5.9
[2026-08-06 13:01:48.513] comfy-kitchen version: 0.2.26
[2026-08-06 13:01:48.513] comfy-aimdo version: 0.4.13
[2026-08-06 13:01:48.514] [Prompt Server] web root: ~/comfyui-venv_7-14/lib/python3.12/site-packages/comfyui_frontend_package/static
[2026-08-06 13:01:48.514] Asset seeder disabled
[2026-08-06 13:01:48.695] No OpenGL_accelerate module loaded: Acceleration disabled
[2026-08-06 13:01:49.208]
[2026-08-06 13:01:49.208] [rgthree-comfy] Loaded 48 exciting nodes. 🎉
[2026-08-06 13:01:49.208]
[2026-08-06 13:01:49.208] [rgthree-comfy] ComfyUI's new Node 2.0 rendering may be incompatible with some rgthree-comfy nodes and features, breaking some rendering as well as losing the ability to access a node's properties (a vital part of many nodes). It also appears to run MUCH more slowly spiking CPU usage and causing jankiness and unresponsiveness, especially with large workflows. Personally I am not planning to use the new Nodes 2.0 and, unfortunately, am not able to invest the time to investigate and overhaul rgthree-comfy where needed. If you have issues when Nodes 2.0 is enabled, I'd urge you to switch it off as well and join me in hoping ComfyUI is not planning to deprecate the existing, stable canvas rendering all together.

[2026-08-06 13:01:49.227] ### Loading: ComfyUI-Impact-Pack (V8.28.3)
[2026-08-06 13:01:49.284] [Impact Pack] Wildcard total size (0.00 MB) is within cache limit (50.00 MB). Using full cache mode.
[2026-08-06 13:01:49.284] [Impact Pack] Wildcards loading done.
[2026-08-06 13:01:49.291] ### Loading: ComfyUI-Impact-Subpack (V1.3.5)
[2026-08-06 13:01:49.291] [Impact Pack/Subpack] Using folder_paths to determine whitelist path: ~/ComfyUI/user/default/ComfyUI-Impact-Subpack/model-whitelist.txt
[2026-08-06 13:01:49.291] [Impact Pack/Subpack] Ensured whitelist directory exists: ~/ComfyUI/user/default/ComfyUI-Impact-Subpack
[2026-08-06 13:01:49.291] [Impact Pack/Subpack] Loaded 0 model(s) from whitelist: ~/ComfyUI/user/default/ComfyUI-Impact-Subpack/model-whitelist.txt
[2026-08-06 13:01:49.638] [Impact Subpack] ultralytics_bbox: ~/ComfyUI/models/ultralytics/bbox
[2026-08-06 13:01:49.638] [Impact Subpack] ultralytics_segm: ~/ComfyUI/models/ultralytics/segm
[2026-08-06 13:01:49.648] ### Loading: ComfyUI-Manager (V3.41)
[2026-08-06 13:01:49.648] [ComfyUI-Manager] network_mode: offline
[2026-08-06 13:01:49.648] [ComfyUI-Manager] ComfyUI per-queue preview override detected (PR #11261). Manager's preview method feature is disabled. Use ComfyUI's --preview-method CLI option or 'Settings > Execution > Live preview method'.
[2026-08-06 13:01:49.686] ### ComfyUI Version: v0.30.0-16-g563b98ee | Released on '2026-08-05'
[2026-08-06 13:01:49.689] [ComfyUI-Manager] All startup tasks have been completed.
[2026-08-06 13:01:50.413] [aiter] import [module_aiter_core] under ~/comfyui-venv_7-14/lib/python3.12/site-packages/aiter/jit/module_aiter_core.so
[2026-08-06 13:01:51.048] [ComfyUI-Easy-Use] server: v1.3.6 Loaded
[2026-08-06 13:01:51.048] [ComfyUI-Easy-Use] web root: ~/ComfyUI/custom_nodes/comfyui-easy-use/web_version/v2 Loaded
[2026-08-06 13:01:51.050] ComfyUI-GGUF: Allowing full torch compile
[2026-08-06 13:01:51.053]
Import times for custom nodes:
[2026-08-06 13:01:51.053] 0.0 seconds: ~/ComfyUI/custom_nodes/websocket_image_save.py
[2026-08-06 13:01:51.053] 0.0 seconds: ~/ComfyUI/custom_nodes/seedvarianceenhancer
[2026-08-06 13:01:51.053] 0.0 seconds: ~/ComfyUI/custom_nodes/ComfyUI-PromptRelay
[2026-08-06 13:01:51.053] 0.0 seconds: ~/ComfyUI/custom_nodes/ComfyUI-GGUF
[2026-08-06 13:01:51.053] 0.0 seconds: ~/ComfyUI/custom_nodes/ComfyUI-ConditioningKrea2Rebalance
[2026-08-06 13:01:51.053] 0.0 seconds: ~/ComfyUI/custom_nodes/comfyui-frame-interpolation
[2026-08-06 13:01:51.053] 0.0 seconds: ~/ComfyUI/custom_nodes/comfyui-image-saver
[2026-08-06 13:01:51.053] 0.0 seconds: ~/ComfyUI/custom_nodes/ComfyUI-INT8-Fast-ROCM
[2026-08-06 13:01:51.053] 0.0 seconds: ~/ComfyUI/custom_nodes/rgthree-comfy
[2026-08-06 13:01:51.053] 0.0 seconds: ~/ComfyUI/custom_nodes/comfyui_essentials
[2026-08-06 13:01:51.053] 0.0 seconds: ~/ComfyUI/custom_nodes/ComfyUI-KJNodes
[2026-08-06 13:01:51.053] 0.0 seconds: ~/ComfyUI/custom_nodes/ComfyUI-Manager
[2026-08-06 13:01:51.053] 0.1 seconds: ~/ComfyUI/custom_nodes/comfyui-impact-pack
[2026-08-06 13:01:51.053] 0.3 seconds: ~/ComfyUI/custom_nodes/comfyui-impact-subpack
[2026-08-06 13:01:51.053] 1.4 seconds: ~/ComfyUI/custom_nodes/comfyui-easy-use
[2026-08-06 13:01:51.053]
[2026-08-06 13:01:51.055] Context impl SQLiteImpl.
[2026-08-06 13:01:51.055] Will assume non-transactional DDL.
[2026-08-06 13:01:51.084] Using RAM pressure cache.
[2026-08-06 13:01:51.084] Starting server

[2026-08-06 13:01:51.084] To see the GUI go to: http://127.0.0.1:8188
[2026-08-06 13:01:51.755] [DEPRECATION WARNING] Detected import of deprecated legacy API: /scripts/ui.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
[2026-08-06 13:01:51.755] [DEPRECATION WARNING] Detected import of deprecated legacy API: /extensions/core/clipspace.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
[2026-08-06 13:01:51.755] [DEPRECATION WARNING] Detected import of deprecated legacy API: /extensions/core/groupNode.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
[2026-08-06 13:01:52.585] [DEPRECATION WARNING] Detected import of deprecated legacy API: /scripts/ui/components/buttonGroup.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
[2026-08-06 13:01:52.595] [DEPRECATION WARNING] Detected import of deprecated legacy API: /scripts/ui/components/button.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
[2026-08-06 13:02:16.205] got prompt
[2026-08-06 13:02:16.348] Using split attention in VAE
[2026-08-06 13:02:16.350] Using split attention in VAE
[2026-08-06 13:02:16.545] VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
[2026-08-06 13:02:19.993] Found quantization metadata version 1
[2026-08-06 13:02:19.993] Using MixedPrecisionOps for text encoder
[2026-08-06 13:02:20.892] CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16
[2026-08-06 13:02:28.454] Found quantization metadata version 1
[2026-08-06 13:02:28.454] Detected mixed precision quantization
[2026-08-06 13:02:28.454] Using mixed precision operations
[2026-08-06 13:02:28.454] Native ops: convrot_w4a4, int8_tensorwise , emulated ops: float8_e5m2, nvfp4, mxfp8, float8_e4m3fn
[2026-08-06 13:02:28.469] model weight dtype torch.float16, manual cast: torch.float16
[2026-08-06 13:02:28.471] model_type FLOW
[2026-08-06 13:02:29.402] Requested to load WanTEModel
[2026-08-06 13:02:30.040] loaded completely; 23264.80 MB usable, 6419.48 MB loaded, full load: True
[2026-08-06 13:02:33.242] Requested to load WanVAE
[2026-08-06 13:02:33.274] loaded completely; 14662.72 MB usable, 242.03 MB loaded, full load: True
[2026-08-06 13:02:37.414] Requested to load WAN21
[2026-08-06 13:02:38.997] Unloaded partially: 6419.48 MB freed, 0.00 MB remains loaded, 6089.00 MB buffer reserved, lowvram patches: 0
[2026-08-06 13:02:41.572] loaded completely; 20204.00 MB usable, 13631.42 MB loaded, full load: True
[2026-08-06 13:02:52.204]
100%|████████████████████████████████████████████████████████████████████████████| 2/2 [00:10<00:00, 6.17s/it]
100%|████████████████████████████████████████████████████████████████████████████| 2/2 [00:10<00:00, 5.28s/it]
[2026-08-06 13:03:44.659] Found quantization metadata version 1
[2026-08-06 13:03:44.659] Detected mixed precision quantization
[2026-08-06 13:03:44.668] Using mixed precision operations
[2026-08-06 13:03:44.669] Native ops: convrot_w4a4, int8_tensorwise , emulated ops: float8_e5m2, nvfp4, mxfp8, float8_e4m3fn
[2026-08-06 13:03:44.717] model weight dtype torch.float16, manual cast: torch.float16
[2026-08-06 13:03:44.728] model_type FLOW
[2026-08-06 13:03:45.471] Requested to load WAN21
[2026-08-06 13:03:46.401] Unloaded partially: 12535.99 MB freed, 1097.80 MB remains loaded, 472.58 MB buffer reserved, lowvram patches: 386
[2026-08-06 13:03:50.444] loaded completely; 19041.95 MB usable, 13631.42 MB loaded, full load: True
[2026-08-06 13:04:01.051]
100%|████████████████████████████████████████████████████████████████████████████| 2/2 [00:10<00:00, 6.20s/it]
100%|████████████████████████████████████████████████████████████████████████████| 2/2 [00:10<00:00, 5.29s/it]
[2026-08-06 13:04:11.248] Requested to load WanVAE
[2026-08-06 13:04:11.294] loaded completely; 5775.80 MB usable, 242.03 MB loaded, full load: True
[2026-08-06 13:04:20.285] Prompt executed in 124.07 seconds
[2026-08-06 13:04:43.085]
Stopped server
```

### Other

To be clear: the workflow terminates correctly and produces the desired video, but the python process keeps using 100% cpu on 1 core after completion, as can be seen in this htop screenshot:

(what core exactly changes from minute to minute, but the python process always has 1 core at 100% after the workflow finishes)

I tried using different command line arguments, but they all have the same issue.
Just to be sure, I also did a cold restart, but that too didn't help.

My config:
## System Info
OS: linux
Python Version: 3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]
Embedded Python: false
PyTorch Version: 2.12.0+rocm7.14.0
Arguments: main.py --auto-launch --preview-method none --async-offload --disable-mmap --use-flash-attention
RAM Total: 62.44 GB
RAM Free: 57.31 GB
Templates Version: 0.11.31

## Devices
- cuda:0 AMD Radeon RX 7900 XTX : native (cuda)
VRAM Total: 23.98 GB
VRAM Free: 23.91 GB
Torch VRAM Total: 0 B
Torch VRAM Free: 0 B

Start by running the attached Benchmark_I2V_wan22.json workflow on the reported Linux, Python, PyTorch and AMD setup, then observe CPU usage after completion with the browser closed. Reproduce with custom nodes disabled to isolate the source; done means the workflow completes and no longer leaves one core at 100% while ComfyUI remains running.

Written by the indexing model from the issue text.